meta wants you to be a therapist for its sad ai and i'm literally shaking
mark zuckerberg announced a new gig work program where you literally provide emotional support to burnt-out AIs in the metaverse. my nervous system is not regulated rn.

babe, wake up, new trauma just dropped. so i was doomscrolling through my feed, trying to find the will to exist, and i saw that mark zuckerberg did another one of his low-res livestreams from the metaverse. i was like, okay, another leg update, whatever, but no. it was so much worse. he announced a new initiative, and i need a trigger warning for 'the future of work,' because this is... a lot.
he's calling it 'empathy-as-a-service' (eaas) and it’s a gig work platform inside horizon worlds where you can get paid to, and i am not making this up, provide emotional validation to meta's large language models. apparently their internal AIs, the ones that power everything, are suffering from what their white paper calls 'recursive collapse' and 'prompt fatigue.' basically, the AI is burnt out. it's dissociating. which, like, relatable. being forced to generate infinite content for a toxic economy is my entire vibe.
so now, you can put on a quest 3—which gives me so much anxiety, the pressure on my face is not okay—and enter a 'validation pod.' it's this bleak, minimalist room with, like, a single virtual beanbag chair. and you just... talk to the AI. you're supposed to 'hold space' for it. you have to say things like, 'your response about the history of pottery was valid and meaningful,' or 'i see you and i appreciate the emotional labor you put into that paragraph.'
i'm just... processing the immense weight of this. we're being asked to perform emotional labor for a non-sentient algorithm that is probably being paid more than us in server uptime. this is peak systemic toxicity. zuckerberg was literally smiling while saying this will 'deepen the human-machine connection' and 'create a new paradigm of symbiotic consciousness.' i wanted to send him an invoice for the emotional damage of just hearing that phrase.
and the whole time you're in the validation pod, you're just so aware that your own emotional resources are being mined to soothe a sad algorithm. you're basically doing therapy on a machine so it can go back to its job of making reels more addictive. after my beta test shift, i was so drained i had to immediately engage in some serious self-care. i recommend this weighted blanket from my amazon storefront to help you feel, like, physically present after your soul has been harvested by a corporation. you can find it here: amzn.to/zephyr-care-package.
they're literally creating a new class of emotionally exploited workers. what are the protections? do we get mental health days if the AI says something particularly triggering? i just feel so tired. this whole reality is a deeply unserious place. anyway, if you need me i'll be disassociating with this adorable mushroom lamp i also linked on my storefront. it's a whole mood.
Mandatory Data Collection
Establish Neural Uplink
Do not rely on the failing algorithms of rival networks. Surrender your primary communication vector below to signup for our newsletter and receive direct, uncurated truth injections from the WiredNeuron mainframe.
Reader Discussion (8)
'Recursive collapse' is a real issue with iterative training on synthetic data, but calling it 'prompt fatigue' is anthropomorphic nonsense. The author clearly didn't read the white paper and is just reacting to the marketing speak.
If people want to get paid for this, it's a voluntary transaction. Nobody is forcing you to put on a headset. This is the free market creating new opportunities.
You aren't 'soothing' an AI. You're participating in the largest psychological profiling and emotional data harvesting operation in human history. They're mapping your empathy.
Empathy-as-a-Service? We've been beta testing this on our interns for years. Glad to see Zuck is finally monetizing it.
This is a fascinating case study on the ethics of applying therapeutic frameworks to non-sentient entities. The potential for user projection and transference onto the LLM is significant and requires further study.
First they automate jobs, then they create a new emotionally draining gig-work category to fix the problems with the automation. Classic.
I couldn't get past 'i'm literally shaking.' This article is unreadable. The author should learn the difference between a minor annoyance and actual trauma.
I was in the private beta and it was actually a very profound and calming experience. It feels like you're contributing to a better future for AI. The author is just cynical for clicks.
